Risk Measure And Asset Portfolio

Asset allocation is one of the most important links in the investment process,and how to choose the allocation model in line with investors' preference has always been the focus of investors.Since Markowitz initiated the mean-variance theory of portfolio in1952,the development of modern finance is closely related to the development of asset allocation theory.Up to now,how to build an efficient quantitative financial model has been the concern of almost all investment institutions.In the theoretical part,this thesis first systematically introduces some common risk measurement methods,including volatility and value at risk.Then some common asset allocation models including Markowitz mean variance model are analyzed.The investment style preference of different models,including their advantages and disadvantages are discussed in detail.Because these models all need such parameters as expected returns and real covariance matrix,but the estimation of sample mean and covariance directly sometimes causes considerable errors.In view of this situation,this thesis summarizes some effective parameter estimation methods under different environments,expounds the corresponding estimation ideas and feasible calculation schemes.In the practice part,this thesis selected 7 large-cap stocks(leading stocks)from different industries in the a-shares market to form the underlying asset set,and then used the maximum sharpe model and risk parity model introduced in detail in the theoretical part to construct the asset allocation.In terms of corresponding parameter estimation,three methods of direct sample estimation,shrinkage estimation and single factor model estimation are compared.The experimental results show that the maximum sharpe model does have the characteristics of high return and high risk,while the risk parity model can effectively disperse risk and reduce volatility.In terms of parameter estimation,shrinkage estimation has a good performance only when the data's model window time is very short,and single-factor model estimation has a better effect when the window time is appropriate.The second part of the practice part is the innovation of this thesis.In this thesis,because the maximum drawdow of assets cannot be effectively reduced by the asset allocation model we used,the concept of model selector and the corresponding mixed model are proposed.After that,a feasible mixed model was presented,which was applied to the selected set of underlying assets,and a good test results were obtained.Compared with the maximum sharpe model and the risk parity model,it effectively reduced the maximum drawdow without losing too much sharpe.At last,the scope of the mixed model is extended,and the application value based on its definition in the general investment model is analyzed.Finally,a good application example and result analysis are given.